README.md

metadata

base_model: deepseek-ai/deepseek-coder-1.3b-base
tags:
  - trl
  - kto
  - generated_from_trainer
model-index:
  - name: k1
    results: []

k1

This model is a fine-tuned version of deepseek-ai/deepseek-coder-1.3b-base on the None dataset. It achieves the following results on the evaluation set:

• Loss: 0.1316
• Eval/rewards/chosen: 2.3534
• Eval/logps/chosen: -119.0631
• Eval/rewards/rejected: -14.8466
• Eval/logps/rejected: -329.5165
• Eval/rewards/margins: 17.1999
• Eval/kl: 2.9675

Model description

More information needed

Intended uses & limitations

More information needed

Training and evaluation data

More information needed

Training procedure

Training hyperparameters

The following hyperparameters were used during training:

• learning_rate: 0.0001
• train_batch_size: 8
• eval_batch_size: 8
• seed: 42
• gradient_accumulation_steps: 16
• total_train_batch_size: 128
• optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
• lr_scheduler_type: linear
• lr_scheduler_warmup_ratio: 0.1
• lr_scheduler_warmup_steps: 200
• num_epochs: 2
• mixed_precision_training: Native AMP

Training results

Training Loss	Epoch	Step	Validation Loss
0.0963	1.7058	100	0.1316	2.9675

Framework versions

• Transformers 4.43.0.dev0
• Pytorch 2.2.2+cu121
• Datasets 2.19.2
• Tokenizers 0.19.1